The Applied Neuroscience Conference was held in Nijmegen, the Netherlands, from May 17-20. I imagine that for many the experience was as singular and as epochal as the first Winter Brain Conference in Key West was for us in 1993. The Society for Applied Neuroscience chose not to have a conference every year, on the grounds that there would not be sufficient new material to justify the effort! But in response to this planned hiatus an ad-hoc organization formed to put on this conference in The Netherlands, which may well have the highest density of neurofeedback practitioners anywhere in the world.

There were some 220 attendees, with 21 nationalities represented, of which 50% were from Holland. Appropriately, the conference was held in Nijmegen, where some 23% of the working population is concerned with either health care or education. It is also a very healthy place to live. The dominant form of transport in this university town is the bicycle. I estimated that there were some 1300 bicycles parked at the train station in dedicated space. One did not see obesity prominently anywhere.

This conference attracted even more of the leading lights in neuroscience than the earlier ones. The conference brought together Niels Birbaumer, Fernando Lopes da Silva, Gert Pfurtscheller, Walter Freeman, Roberto Pascual-Marqui, Evian Gordon, E. Roy John, and Juri Kropotov as keynoters.

Niels Birbaumer gave the first keynote address, and he really covered the waterfront on Brain-Computer Interfaces (BCI). His group is now including neurofeedback in general within the term BCI, even though neurofeedback does not have the historical baggage in Europe that it does in the United States. Birbaumer has always had to be strategic in his thinking. His work on locked-in-syndrome over the years has had to be sold as research rather than as part of treatment throughout. When competing in a medical setting as a non-MD, he suggested, one also has to come with bigger equipment, which in this instance means an fMRI imager. And now the training paradigm in neurofeedback has a chance to hitch a ride on the more exotic, invasive techniques being evaluated for BCI (provided of course that it is accompanied by fMRI measurements…).

Birbaumer recounted some of his work with slow cortical potentials (SCPs), and in particular recalled an encounter with an irritating neurofeedback skeptic some years ago. Since the fellow did not take the work seriously, he allowed himself to be hooked up to the instrument and to be trained toward higher cortical excitability. “If you do this, you will get seizures,” Birbaumer told him. When the fellow had his first-ever seizure right then and there, he adjusted his thinking…. In the real world, it’s not controlled studies that matter; it’s whether you can predict outcome.

Birbaumer pointed out that people who learn to control the SCP can still perform the task a year later, and possibly even exceed their earlier performance. The prominent lack of an extinction issue in neurofeedback suggests homology with motor skill learning.

And if a motor skill model applies, how important is the issue of conscious perception? Trainees tend to report the use of cognitive strategies to achieve the goals of SCP training, but in fact the acquisition of cortical control typically precedes the perception of control of SCP, and sometimes that perception never materializes at all. The critical variable is the contingency of the cortical response. Much of this whole process can be understood at the brain level.

Birbaumer made a big point of the fact that these folks–locked-in patients, ALS patients, etc.–manifested a quite satisfactory quality of life, better in fact than that of their own doctors and family members. This seemed to hinge on the ability to communicate that the SCP training was facilitating. “The human being lives according to its communication capacity: losing the capacity for communication means losing life.” These people don’t die! It may be the capacity to communicate that is keeping them from slipping into the completely locked-in state and is keeping them alive. The state of their psychological health may even improve over time. Birbaumer implicitly chided the Dutch because in Holland euthanasia is legal, and locked-in patients are taken off life-support: “Don’t kill these people,” he pleaded.

When training was attempted with totally locked-in patients (no eye or eye-lid movements; no sphincter control) the attempt was often futile. Maintaining a communications capability was one thing; establishing such an ability in the first place may be quite another. A deeper problem is indicated, and Birbaumer suggested that once the capacity for intentional agency is lost, a kind of extinction process may get underway that forestalls any recovery. “If you remain in the dark, eventually you lose the capacity to see.” “If you undergo extinction for goal-directed thinking…. you give up–not emotionally–but rather you give up goal-directed thinking.”

At this point in his talk I digressed to thinking about the severely cognitively impaired children that we have trained successfully. These were children who manifested no cognitive awareness of their surroundings, and yet they responded to the neurofeedback challenge. The whole process had to be explainable in brain terms, and at the brain level. Not so with Birbaumer’s work: a minimum level of consciousness was required to sustain this process because intention needed to be mobilized, on the one hand, and communication was the objective, on the other.

Our neurofeedback training admits of a more modular conception: Spasticity should be trainable if only the cortex is able to detect a correlation of the feedback signal with its own state. We require nothing more than that the brain of the child retain “state awareness,” on the one hand, and some level of capacity to process incoming sensory information, on the other. In particular, we do not need to postulate higher-level cognitive processes such as the mobilization of intention. For the rest of the story, nature simply takes its course, which is that the brain seeks correlations or, to put it less anthropomorphically, the persistence of correlations creates attractor states for the brain.

The difference in “accessibility” of the two methods is most likely attributable to the fact that in frequency-based neurofeedback we are speaking in the brain’s natural language. For as long as the brain consumes oxygen it is going to be organizing its activities in the frequency domain and resisting arbitrary intrusion into its affairs. This is a world away from asking a person to change his slow cortical potential at a particular scalp location by at least four microvolts following a prompt.

In the work with ALS patients, trainers look for any responsive muscle group and use that for training purposes. They also check cortical responses to assure that the patients are not in coma, and that cognition is sufficient to support a learning paradigm. For this to work there must be substantial coordination between brain modules, and there must be some higher-level competences. The floor is therefore a lot higher than it is for frequency-based neurofeedback.

It was at this point that Birbaumer reflected on the early work of Neal Miller, who was engaged on the topic of whether autonomic response training was contingent on a muscular response. The very question might strike the present-day neurofeedback practitioner as bizarre, but in the sixties “the belief that instrumental learning is possible only for the cerebrospinal system and, conversely, that the autonomic nervous system can be modified only by classical conditioning” [Neil Miller] was firmly established, even though the evidence was weak. Said Miller, “It is only against a cultural background of great prejudice that such weak evidence could lead to such a strong conviction….” (Neal Miller, Learning of Visceral and Glandular Responses.)

This led Miller to train autonomic function via operant conditioning techniques in curarized rats. The results were epoch-making at the time. Subsequent replication repeatedly failed, however, one consequence of which was the death by suicide of one of Miller’s former graduate students who felt disgraced by this affair. Miller was essentially correct, however, in questioning the hard separation between CNS regulatory functions, on the one hand, and conditioning modalities, on the other. Ironically, this early compartmentalization of thinking carries over to the present day, even within the biofeedback community itself, with traditional biofeedback people still incredulous at the suggestion that the EEG is a perfectly suitable mediator for training autonomic nervous system regulation.

The story is relevant also because of Niels Birbaumer’s personal biography. He appeared on the scene at Rockefeller University in 1966 just in time to be involved in the fateful replication of Miller’s work….

Birbaumer also covered the waterfront on other kinds of neurofeedback in his talk. He spoke favorably of the ability to train down chronic pain using fMRI as a measure even within a single session (the work of deCharms). Of course we routinely do the same with EEG training, and this work is covered in a recent publication (Jensen et al, 2007). Almost inevitably scientists allow themselves to be beguiled by evidence of specificity even when the proposition remains to be proved. Chronic pain is almost surely not localizable, but rather reflects the “state of the system,” or more specifically, the state of the networks. Moreover, it is clearly responsive to methods that do not make any assumptions about localization.

Birbaumer then turned his attentions to the emotions, quoting Gustav Mahler:

“Die einzige Wahrheit auf der Erde ist unser Gefühl”

“The only truth on earth is our feeling,”
which both updated Descartes and anticipated Antonio Damasio’s “Descartes Error.”

“Cortex is useless for the emotions,” Niels declared. Of course this stubborn reality does not diminish the zeal of those who look to the QEEG for guidance in training emotional regulation in the autism spectrum and in Conduct Disorder! To Niels this meant that one needed direct access to the subcortical centers, following deCharms. In working with psychopaths, Birbaumer showed that emotional responsiveness of the insula region could be substantially enhanced in a mere three hours in the fMRI scanner. Well, that’s still cortex, but let’s not quibble.

Despite the positive evidence of change in insula responsiveness to emotionally charged, particularly negative, stimuli, Birbaumer was still not sure whether his finding translates into behavioral normalization as well. But by now we already have evidence accruing from a number of sources to the effect that we do influence psychopathy favorably with neurofeedback. For example, we see it in corporate executives who have managed to escape the label, but who clearly fit the pattern.

The reverse is also true, however. Just now we are hearing about a child with Reactive Attachment Disorder being turned into a budding sociopath in only a single session of inappropriate EEG training. This phenomenon has been known for years. So, what then is the truth of the matter? Is cortex involved in emotional regulation or not? It is again a matter of the state of the system, to which the EEG bears witness. The EEG does give us sufficient information to train emotional regulation based on the EEG alone. The bumps and valleys in the QEEG, however, are likely to reflect other issues. A psychopath can have a perfectly normal EEG in every way… There is no EEG signature for psychopathy. And yet any psychopath will respond as readily to EEG neurofeedback as anyone else. And the response will be just as rapid as it is with fMRI-based training.

Birbaumer then turned to near infrared training as a possible means of training infants in utero. The fetus is bathed in high-intensity infrared signal, and the return signal is monitored. Another possibility is to do magnetoencephalography (MEG) on the fetal brain. But the obvious pathway to training the fetal brain is to train the mother, and that has already been done to great effect in one very dramatic case. The two nervous systems are intimately coupled, albeit non-synaptically. Again one is moved to ask, why not talk to the clinicians first? In this tortoise-and-hare situation, every time the hare arrives at his goal, he will find the tortoise, the clinician, already there.

At the end of the talk, Birbaumer took a rather strong position in favor of neurofeedback, and in this paean he clearly included the frequency-based EEG training and the other nascent technologies along with his own slow-cortical potential training:

“Behavioral learning treatments based on solid neurobiological evidence constitute the most effective approach to the treatment and rehabilitation of many neurological and psychiatric diseases. Acceptance and clinical use of these strategies is negligible because of a lack of anticipated profit, and a lack of interdisciplinarity….”

A fellow attendee remarked that this fervent message was actually somewhat unprecedented for Birbaumer. The emphasis earlier had always been on the communication issue with the locked-in patients, where functional remediation was not really in prospect. No controversy ever attached to that work. The fact of communication was its own sufficient proof. Now Birbaumer was clearly wading into deeper waters. In fact, he was moved to recall an old Yiddish saying,

“Fun lojter hofenung wer ich noch meschuge…”
“Vor lauter Hoffnung werd’ ich noch verrückt….”
“I have so much hope that it drives me batty…”

I had hoped to cover the whole conference with this newsletter, and instead I have only covered Birbaumer’s talk. In some sense this is appropriate, in that Birbaumer’s career spans the entire history of EEG neurofeedback as very few others do. Every other speaker at the conference is a comparative latecomer to the cause, and none of the other keynote addresses even focused on neurofeedback. Now Birbaumer is moving adroitly toward a much-needed integrative perspective along with being willing to take an unapologetic stance with respect to what we collectively have to offer the world.

The following day, over morning coffee, I heard a graduate student attendee wonder openly why there was so much neurofeedback at this conference. She was obviously unaware that neurofeedback had provided the original impetus for the conference among the planners. I, on the other hand, had wondered why there was so little neurofeedback on the program. Throughout the conference I returned to the thought that surely the proceedings would be very different if the various researchers really knew what we routinely accomplish clinically.

The objective of the planners is clear. It is to embed the wayward stepchild of neurofeedback within enough classical neuroscience that it acquires legitimacy by association. I am reminded of men’s clothing stores of an earlier generation where the salesmen were all out front in spiffy garb, but the real professional on the scene was the tailor in the back room, who came to his task in work clothes, with sleeves rolled up, and with his lips clamping down on a bunch of stickpins. He was the guy who really knew the business, and who was ultimately responsible for customer satisfaction. The clinician similarly has the task of making every client visit productive.

So we have the analogy that the neuroscientists serve as the suits that bring respectability to the enterprise. All the while the real business is making people well, and we keep that more complex, less pristine and hence less “scientific” part of the business under wraps.

When it comes to other leading technologies, genomics and stem cell research, the great clinical potential is the obvious driver. But, when it comes to neurofeedback, the practical application always gets second billing. Jonathan Walker was in the audience, for example, but there was no exposure to his latest work with seizure disorder using coherence-based training. There was no coverage at all of coherence-based training, or of stimulation-based training; there was nothing on the autism spectrum. My own talk was not scheduled until the last afternoon, by which time Birbaumer had left.

Sadly, this was an opportunity missed because our own work is migrating to the point where it offers a significant point of contact with Birbaumer’s SCP work. Much of our training is now occurring in the frequency domain of 0.1Hz. The difference is that when we train there, we are staying entirely within the frequency domain of slow cortical potentials. This is not the case, for example, with the standard SCP training. If one induces a transient in the SCP, a spectral analysis of such a transient clearly takes us beyond the SCP domain. What processes are involved in the induction of such a transient? We don’t really know. By contrast, when the only information presented to the trainee relates to 0.1-Hz activity, we are training unambiguously within the bounds of the SCP regime. A contrast can therefore be drawn both with the more conventional higher-frequency training and with Birbaumer’s SCP training procedure.

By working in the frequency domain we also get to take advantage of the efficiency of frequency-based training that has already been alluded to above. The demands on the client are reduced, and we utilize 100% of the timeline to inform the feedback. The movement to these low frequencies has improved our training efficiency with the relevant subset of clients who need to train there. It would be nice to see a replication of our work by the Tübingen group.

Much of the subsequent discussion at the conference was about how to improve neurofeedback by cleaning up the data stream with such techniques as Independent Component Analysis (ICA), LORETA analysis (sLORETA), etc. It is as if neurofeedback needed all the help it can get. The standards for this work are set by the needs of EEG measurement and assessment, however. What is not appreciated by the research community is that the bar for signal integrity is much higher in the realm of assessment and characterization than it is in neurofeedback training itself.

The difference may be illustrated with another anecdote. One of the essential messages of the work of Walter Freeman is that the brain makes rather economical use of the information provided to its sensory processing apparatus. In a chat one evening he brought up the recent experiment in which the output of a videocam was mapped to the tongue of a blind person. Only 24 “sensors” were involved. The experimental subject was able to reconstruct a visual representation out of that sparse signal stream, one that was sufficient to allow him to navigate through space. Cochlear implants don’t allow for a large number of connections either, yet they allow a fair reconstruction of one’s auditory environment.

In neurofeedback, matters are even simpler. The “information space” that is being represented in neurofeedback is one that the brain is already acquainted with. All the brain has to do is to recognize and track a correlation between the proffered signal stream and the goings-on within its own environs. When I suggested this model to Freeman, who is still out in the vestibule on neurofeedback as far as I can tell, he responded that he preferred not to use the word “information” with regard to the brain. It sends the wrong message. If we think in terms of information retention, even our large brains would be quickly overtaxed. The brain clearly retains something more like “knowledge,” or some such abstraction above and beyond the level of mere information. Just as the brain is economical in its dealings with sensory information, it must be economical in what it retains. Of course it is not so much an issue here of what the brain retains over time. Rather, it is a question of what level of awareness one may attribute to the brain in the moment. For neurofeedback to work, the brain must at minimum have “state awareness.”

Just as we have a much more profound experience of the present moment than we do of any past moment, the same must be true of the brain with respect to its own processes. Now the kind of awareness postulated here is not the one we associate with consciousness. Rather, we need to be able to explain the empirical observation that the brain may respond to neurofeedback in a location-specific and frequency-specific way within a minute or less of a change in protocol. In order for this to be possible, the brain must be engaging in a continuous pattern-matching activity with respect to the incoming signal stream.

The question then becomes how noise-immune this process is. Our experience tells us that the noise immunity is impressive, however one looks at it. We feed the brain a fragment of a complex signal, one that is adulterated by various artifacts, and yet the brain is able to sort things out. But this capacity is not novel. A similar discrimination task is accomplished in our visual system. If we sit in a Land Rover bounding over the terrain in the Kalahari Desert, we still see the world beyond in terms of a stable image. Our sidekick the cameraman, however, is having a hard time of it with his old video gear, devoid of an image stabilizer. When folks look at his imagery, taken from the moving Land Rover, they immediately get nauseous. When we do neurofeedback, we benefit from the same tricks our visual system uses. The brain is aware when eyes move, and it is similarly aware of its own artifacts. When it comes to EEG analysis, however, we have the same problems as folks have when they watch the cameraman’s video imagery.

Just as frequency-based neurofeedback is less demanding on the “system” than standard SCP-training, frequency-based brain stimulation is simpler and more straight-forward than traditional neurofeedback (which is not to say that it is preferable clinically). If the brain is exposed to a steady stimulation within the EEG band, the brain will necessarily respond. A frequency-specific state shift may be produced almost immediately. The stimulation is an intrusion into the brain’s frequency space that is as impossible to ignore as it would be for us to ignore someone starting to play the flute in our sound space. As Wordsworth said, “The eye cannot help but see, nor the ear to hear.” This stimulation is occurring entirely within the brain’s domain, and can therefore be considered a passive technique.

With traditional neurofeedback, we depend upon an additional critical step, the detection of a correlation of the “abstract” feedback signal with what is going on within the brain. And then the brain must mobilize a response. Nothing should alter the frequency characteristics of the EEG or the state of the system unless those two steps occur. It is actually quite remarkable that this works as well as it does, but of that there is no longer any doubt. What is repetitively practiced is also learned, and the brain’s functional domain expands.

“I am mad with hope…”

Siegfried Othmer

Reference:
Neurofeedback treatment for pain associated with Complex Regional Pain Syndrome Type I, Mark P. Jensen Ph.D., Caroline Grierson, R.N., Veronika Tracy-Smith, Ph.D., Stacy C. Bacigalupi, M.A., Siegfried Othmer, Ph.D.

Journal of Neurotherapy, 11(1), pp 45-53 (2007)

This entry was posted on Monday, June 18th, 2007 at 6:46 am and is filed under Conferences, Neurofeedback, Research. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.